WorkBuddy / CodeBuddy

WorkBuddy / CodeBuddy are AI agent and coding-assistant tools from Tencent Cloud. They support adding custom models via a local model config file, so you can connect NovaAPI through the OpenAI-compatible Chat Completions API.

1. Install and sign in

• Download and install the WorkBuddy / CodeBuddy desktop app, or install the command-line edition (CodeBuddy Code CLI):

  npm install -g @tencent-ai/codebuddy-code

• Launch it once and complete sign-in as prompted.

• Open a project directory at least once so the app creates its local config directory (.codebuddy).

• Get an API Key from the NovaAPI console (starts with sk-).

2. Configure a local model

Create or edit the user-level config file:

• Windows: C:\Users\<your-username>\.codebuddy\models.json
• macOS / Linux: ~/.codebuddy/models.json

To scope the config to a single project, create a project-level file instead: <your-project>/.codebuddy/models.json.

First set your NovaAPI Key as an environment variable:

# Windows PowerShell (persistent, reopen terminal)
setx NOVA_API_KEY "sk-your-NovaAPI-key"

# macOS / Linux
export NOVA_API_KEY="sk-your-NovaAPI-key"

Then write this config:

{
  "models": [
    {
      "id": "gpt-5",
      "name": "GPT-5",
      "vendor": "NovaAPI",
      "url": "https://api.novaapis.com/v1/chat/completions",
      "apiKey": "${NOVA_API_KEY}",
      "maxInputTokens": 128000,
      "maxOutputTokens": 8192,
      "supportsToolCall": true,
      "supportsImages": false
    }
  ],
  "availableModels": ["gpt-5"]
}

• url: NovaAPI's OpenAI-compatible endpoint — include the full /v1/chat/completions path
• apiKey: ${NOVA_API_KEY} reads the key from an env var (you can also paste the sk-... key directly)
• id / name: the model ID and display name; take id from the model list, add more entries as needed
• supportsToolCall: set true when using a tool-calling model

Save models.json as UTF-8 without BOM. Some desktop builds fail to load the local model config from a JSON file that carries a UTF-8 BOM.

3. Restart and pick the model

Fully quit WorkBuddy / CodeBuddy, reopen it, and select your configured model (e.g. GPT-5) in the model picker.

4. Optional: verify the API Key

# Windows PowerShell
$env:NOVA_API_KEY="sk-your-NovaAPI-key"
curl https://api.novaapis.com/v1/chat/completions `
  -H "Content-Type: application/json" `
  -H "Authorization: Bearer $env:NOVA_API_KEY" `
  -d '{"model":"gpt-5","messages":[{"role":"user","content":"hi"}],"stream":false}'

# macOS / Linux
curl https://api.novaapis.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $NOVA_API_KEY" \
  -d '{"model":"gpt-5","messages":[{"role":"user","content":"hi"}],"stream":false}'

A successful response confirms the API Key and model name both work.

FAQ

• Authentication fails / 401: check that apiKey is a real NovaAPI Key — don't put the endpoint URL in the API Key field.
• Model not found / 404: check that the model id matches the model list exactly.
• Failed to read local model config: verify models.json is valid JSON and saved as UTF-8 without BOM.
• Not shown in the model picker: fully restart the app and confirm the file is at .codebuddy/models.json.
• UI shows ${NOVA_API_KEY} literally: restart the app from a terminal where the env var is set; if the desktop build still doesn't expand it, put the real API Key directly in the UI or models.json.